Search

Summary: Background: Patients with atrial fibrillation (AF) frequently experience multimorbidity. Cluster analysis, a machine learning method for classifying patients with similar phenotypes, has not yet been used in South Asian AF patients. Methods: The Kerala Atrial Fibrillation Registry is a prospective multicentre cohort study in Kerala, India, and the largest prospective AF registry in South Asia. Hierarchical clustering was used to identify different phenotypic clusters. Outcomes were all-cause mortality, major adverse cardiovascular events (MACE), and composite bleeding events within one-year follow-up. Findings: 3348 patients were included (median age 65.0 [56.0–74.0] years; 48.8% male; median CHA2DS2-VASc 3.0 [2.0–4.0]). Five clusters were identified. Cluster 1: patients aged ≤65 years with rheumatic conditions; Cluster 2: patients aged >65 years with multi-comorbidities, suggestive of cardiovascular-kidney-metabolic syndrome; Cluster 3: patients aged ≤65 years with fewer comorbidities; Cluster 4: heart failure patients with multiple comorbidities; Cluster 5: male patients with lifestyle-related risk factors. Cluster 1, 2 and 4 had significantly higher MACE risk compared to Cluster 3 (Cluster 1: OR 1.36, 95% CI 1.08–1.71; Cluster 2: OR 1.79, 95% CI 1.42–2.25; Cluster 4: OR 1.76, 95% CI 1.31–2.36). The results for other outcomes were similar. Atrial fibrillation Better Care (ABC) pathway in the whole cohort was low (10.1%), especially in Cluster 4 (1.9%). Overall adherence to the ABC pathway was associated with reduced all-cause mortality (OR 0.26, 95% CI 0.15–0.46) and MACE (OR 0.45, 95% CI 0.31–0.46), similar trends were evident in different clusters. Interpretation: Cluster analysis identified distinct phenotypes with implications for outcomes. There was poor ABC pathway adherence overall, but adherence to such integrated care was associated with improved outcomes. Funding: Kerala Chapter of Cardiological Society of India.